Speed and Altitude: The Secrets Revealed for Very Low Altitude Image Processing
It is always interesting, but also challenging, to fly an UAV at very low altitude like 10-20 meter over crops because this gives very fine quality images that not only look very crisp but also act as a very useful asset for ground cover analysis and determination of plant density. In 2015, we conducted such flights over maize with an altitude of 15 meters. The captured images looked fine but while mosaicking the images, the success rate was only 50%. So, this year we took the challenge to solve the problem. We found that it was the small forward overlap for which some images could not be processed. We know that at least 60% overlap is necessary for a UAV captured image mosaic but the flight planning software we are using only deals with lateral/side overlap. We also understood that the lower the altitude, the higher the clarity for ground cover. So this time we focused on a combination of sufficient forward overlap and the lowest possible altitude.
To solve the problem of insufficient forward overlap, we focused on the hardware. The UAV itself has a capability to fly between a speed range of 0.5 m/s to 8 m/s. We observed that the lower the speed, the higher the forward overlap due to the increase in number of pictures per distance covered. Last year, the speed of the UAV was 6m/s which resulted in a forward overlap of 35-40%. This year we lowered the speed to 3m/s which resulted in a 60-70% forward overlap (Fig.1). This tremendous increase in forward overlap ensures success in image processing without any problem.
Forward Overlap at 6m/s
Forward Overlap at 3m/s
Fig.1: Increase forward overlap with the decrease of UAV speed
To increase the resolution, we decided to reduce the flying altitude. This time we flew the UAV at 12 meter high. The lowest altitude our UAV should be flown is 10 meter, but we flew it at 12 meter to be on the safe side. The processed images illustrate that images captured at 12 meter insure better quality than 15 meter (Fig.2). Some credit for the improved image quality goes to lower speed as well. An image taken at 15 meter has a pixel size of 3 mm. Now with the image taken at 12 meter, the pixel size is 2 mm. This better image quality helps to identify ground cover more accurately.
Fig.2: Quality of image taken from 12 meter altitude
Individual plants are clearly visible which makes it possible to count the maize leaves even in the early stage (Fig.3a). With these flight parameters, we are able to cover almost half a hectare in a single flight (Fig.3b).
Fig. 3a: Detailed view of a maize plant, captured with a Sony NEX-7 from 12 m altitude.
- Fig 3b: Area of our maize research plots, which was imaged at a resolution of 2 mm, resulting in 1805 million pixels.
So, with this reduction in speed and altitude, we can produce high quality images. As some risks come with this reduction in speed and altitude as well, the following points should be taken into account. It should be remembered that lower speed costs more battery life to cover a certain area. This implicates that there must be a distinct matching between speed and battery life time otherwise the UAV can’t cover the whole site. Another point to remember is to be careful about the surface clearance and the threshold limit of flight altitude while making flight plan for very low altitude like 12 meter. If these recommendations are taking into account, flying at low altitude can be very rewarding.